PICCA Frederic-Emmanuel <frederic-emmanuel.pi...@synchrotron-soleil.fr> writes:
>> That is mostly upstream's job -- ICD packagers should just verify that the >> package still runs "Hello World" on their hardware, i.e. the ICD >> integration works, and then we assume that it works. > > ok, so in that case it would be nice to provide a computer with a GPU as > porterbox to test this hello world. > > Since we are using a lot of autopkgtest, this should be available for these > integrations tests. +1 because if we don't trust upstream-provided file.texi and are supposed to regenerate it from upstream's markdown files rather than trusting upstream's intermediary format, then how much moreso for something of importance like this! Given our reproduciblity and autopkgtest QA-related goals, we ought to QA the GPU-accelerated information processing on real hardware, because this provides an additional level of trust/confidence for people using Debian in their work and research. Finally, while it is proprietary software, it would be nice if such QA could encourage Davinci Resolve to support to support AMD on Debian. Currently they only support nVidia on RHEL. In other words, I absolutely affirm Mo Zhou's earlier proposal to provide a credible alternative to nVidia's CUDA. (email subject: rocm-team: in support of AMD's free CUDA counterpart and end NVIDIA monopoly) When framed this way (full-stack high-quality QA), I am optimistic that AMD would sponsor Debian--replying to Sam Hartman, who wrote: If AMD wants to sponsor Debian by giving us some hardware, I definitely think we should see if we can make that work. Cheers, Nicholas
signature.asc
Description: PGP signature
